Kohn discusses the design and implementation of an architecture for an intelligent real-time control system. The first section of the paper describes the difference between a conventional controller and a declarative controller, establishes the need for an intelligent formal method for designing and implementing a control system, and illustrates the steps involved in designing a declarative controller.
The elements of the declarative control architecture are covered next. This architecture is composed of a knowledge base, a theorem planner, an inferencer, and an adapter. The knowledge base uses the first-order logic clauses with second-order extensions for representing knowledge. The denotational semantics of each clause in the knowledge base are based on conservation, invariance, or constraint principles. The theorem planner generates, for each update interval, a statement that describes the desired system behavior during that interval. The inferencer investigates whether the theorem derived by the theorem planner follows logically from the knowledge stored in the knowledge base. The last component of the architecture (the adapter) is not discussed in the paper.
Next, the author describes a prototype implementation of the architecture. He uses an intelligent controller for running a chemical reactor to illustrate the operational and functional characteristics of this implementation. I like the fact that, in implementing the system, the author has emphasized self-awareness and hierarchical decomposition of the inference procedure.
Kohn concludes by describing the system’s current status, conclusions, and prospects for future research. Even though he has stated that he used two other prototypes to illustrate the feasibility of the concept, he never discusses these studies. I would have liked to read about the author’s experience with these prototypes. I would also have liked to know how these prototypes paved the way for the third prototype. What criteria were used to determine the concept’s feasibility? A brief comparison of these prototypes would have paved the way for discussing the third prototype, and would have strengthened the case for implementing the third prototype.